Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Chapter 4<br />
<strong>Nonlinear</strong> Modeling <strong>of</strong> Complex<br />
Waveforms in the f − x Domain<br />
4.1 Linear <strong>Prediction</strong> in the f − x Domain<br />
F - X deconvolution is a popular noise attenuation tool first introduced by Canales<br />
(1984). Seismic traces are represented in the time-space domain. When one trans-<br />
forms each trace into a Fourier domain, the complex waveforms are represented in<br />
the frequency-space domain (f − x). Linear events are predicted for each frequency<br />
in the spatial direction <strong>of</strong> a given frequency. Therefore, if a signal can be predicted,<br />
the difference between the observed and predicted signals can be considered an<br />
estimation <strong>of</strong> the noise in the data.<br />
Linear prediction filters that map to monochromatic complex sinusoids in the f−<br />
x domain can accurately predict linear events in the t − x domain. A superposition<br />
<strong>of</strong> complex harmonics immersed in white noise can be predicted using an ARMA<br />
(autoregressive-moving average) model as suggested by Sacchi and Kuehl (2001).<br />
47